RESEARCH a Riparian Wildlife Habitat

RESEARCH DOI: 10.1007/s002670010249 A Riparian Wildlife Habitat Evaluation Scheme Developed Using GIS

LOUIS R. IVERSON* ual streams (for acquisition or management) by habitat rank- USDA Forest Service ing. We demonstrate this methodology for the Vermilion River Northeastern Research Station basin in east-central Illinois, USA. Three data sets were used 359 Main Road to evaluate land cover encompassing 300 m on either side of Delaware, Ohio 43015, USA the streams: (1) the US Geological Survey’s land use and land cover information (LUDA), (2) land cover manually digitized DIANE L. SZAFONI from the National High Altitude Photography (NHAP) program, Geographic Modeling Spatial Laboratory and (3) Landsat Thematic Mapper (TM) data classiﬁed into University of Illinois at Urbana-Champaign land cover. Each of 30 tributaries in the study area was 220 Davenport Hall ranked for habitat according to the data contained in each Urbana, Illinois 61801, USA data set, and results were compared. Habitat ranking schemes were devised and analysis performed for three spe- SHARON E. BAUM cies guilds: forest, grassland, and mixed successional spe- Illinois Natural History Survey cies. TM and NHAP each differentiated habitat scores (for for- 607 East Peabody est, grassland, and mixed successional guilds) among Champaign, Illinois 61820, USA tributaries in a similar and suitable way, while LUDA was not suitable, due to the coarse resolution of the data. Overall, it ELIZABETH A. COOK was shown that the methodology is suitable to rank streams USDA-Natural Resources Conservation Service at Lincoln based on riparian habitat quality. Even though more work is University needed to test and verify the method, the project has shown 306 Founders Hall the potential for such techniques to assist in evaluating, track- Jefferson City, Missouri 65102, USA ing, and improving the management of riparian wildlife resources. The method can easily be applied over large areas ABSTRACT / To evaluate riparian habitat for wildlife, we used such as states if TM-based land cover and stream data are a geographic information system (GIS) that prioritized individ- available.

In the late 1970s, the President’s Council on Envi- (Hanson et al. 1990, Harris and Scheck 1991), sinks, ronmental Quality (1978) estimated that as much as and sources. 70% of riparian ecosystems (systems adjacent to Illinois is a highly developed state with less than 11% streams and rivers) present at the time of European of its area in its “potential” vegetation type (Klopatek et colonization in the United States had been destroyed. al. 1979). Although large in size (142,000 km2), agri- Presently, evidence continues to mount on the ex- culture and urbanization dominate the state, and only tremely high value that riparian systems provide in a small percentage of the state can be considered high ecosystem services (Forman 1995); protection of water quality for wildlife (Iverson et al. 1989). Illinois con- quality as ﬁlters (Welsch 1991, Gilliam 1994), biodiver- tains in excess of 21,200 linear km of streams and rivers sity/habitat (Naiman et al. 1993), conduits for dispersal (Neely and Heister 1987). The majority is characterized by their low gradient and was historically connected to

KEY WORDS: Landscape ecology; Riparian habitat; Wildlife habitat; expansive floodplain areas of high-quality wetland, for- GIS; Illinois; Spatial analysis est, and grassland habitats. Subsequent channelization, artificial draining, and leveeing effectively isolated the *Author to whom correspondence should be addressed; email: fertile floodplain from the stream channels leading to [email protected]. the decimation of much of the native habitats through- 1 The use of trade, firm, or corporation names in this publication is for out the state (e.g., Osborne et al. 1991). Similar, or the information and convenience of the reader. Such use does not constitute an official endorsement or approval by the US Department even greater losses of forests and wetlands have been of Agriculture or the Forest Service of any product or service to the documented across Illinois (Iverson and Risser 1987; exclusion of others that may be suitable. Iverson et al. 1989, Havera and Suloway 1994), and land

conversion continues, especially in the south-central scribed here. By incorporating a spatial component portion of the state (Osborne et al. 1991, Iverson 1994). into the data sets, the IGIS made it possible to combine The loss of wildlife habitat has been documented in and compare various data sets, evaluate the character- Illinois and has reached a critical stage (Illinois Wildlife ization of streams relative to wildlife habitat, and gen- Habitat Commission 1985, Havera and Suloway 1994). erate maps and tabular data for statistical analysis and Remaining riparian areas comprise a significant por- display. Several data sets were available for analysis and tion of the remnant forested vegetation in the state. In evaluation (Table 1). fact, for the south-central portion of the state, 79% of Many of the data sets used in the study are subsets of the forest land is within 300 m of the streams (Iverson larger, statewide coverages. Others are regional, cover- 1994). This pattern is primarily due to residuals remaining only the Vermilion basin, as in this investigation. ing because of the historical difficulty of growing row Data sets differ in scale, temporality, and in the degree crops economically on the steeper slopes associated of processing that is necessary to use them. All images with many of the stream and river valleys (Iverson (photographic or satellite) were based on data col- 1988). Thus, information on the location and extent of lected within a decade, 1978–1988, to minimize tem- these important vestiges of native Illinois is critical for poral variation. This basin is largely rural and is not the successful management and protection of riparian undergoing rapid urbanization relative to some other habitats. In Illinois, this is complicated not only by the parts of the state and country. Different coding state’s size, but also by the geographical variability that schemes were normalized so that data generated exists from north to south, and most importantly, by through spatial processing of the various data sets could economic constraints. It is imperative that resource be statistically compared. managers maximize the use and application of infor- Land and Stream Data Set Description and mation that may reside in statewide databases to iden- Derivation tify these vital riparian habitats. The objectives of this study were to: (1) identify riparian wildlife habitats Digital line graph. Primary among the digital carto- along streams in the Vermilion River basin located in graphic products produced by the US Geological Ser- east-central Illinois using applicable and available data vice (USGS) are the digital line graph (DLG) files sources; (2) develop and assess methodologies to iden- (Table 1). The 1:100,000-scale hydrographic layer of tify riparian habitats of different qualities in the Ver- the DLG files for all of Illinois resides on the IGIS and milion River basin using these data sources; and (3) comprises the most detailed representation of the compare the results of these methods to identify quality streams and lakes available for the whole state. The riparian wildlife habitats throughout the state and be- USGS produced these data by digitizing a photograph- yond. ically reduced composite of the hydrographic layer from the 1:24,000 USGS base map series. The data carry codes identifying the type of feature represented, such Materials and Methods as stream or shoreline and descriptive information such as flow (for example, intermittent) or spatial location Study Area (for example, right bank) (US Department of the In- This study was conducted on the Salt Fork and Mid- terior and US Geological Survey 1989). dle Fork branches of the Vermilion River (tributary to The data pertinent to the study area were extracted the Wabash River) located in east-central Illinois (Fig- and modified for use in this project. This effort in- ure 1). Within these watersheds, a total of 30 streams volved edge matching to make the data topologically drain an area of at least 10 sq mi (25.8 km2). The consistent and then extracting only those streams with Vermilion River basin has a long history of scientific at least 10 square miles of drainage area. investigation because of its close proximity to the Uni- Land use and data analysis. The USGS Land Use and versity of Illinois and the Illinois Natural History Survey. Data Analysis (LUDA) data set contained within the Many detailed descriptions of the history, geography, IGIS represents land use and land cover for all of soils, climate, land use, and aquatic fauna have been Illinois. LUDA is based on a hierarchical classification reported (for example, Osborne and Wiley 1988; Wiley system developed nationally by the USGS for use with et al. 1990; Osborne et al. 1991; Tazik et al. 1991). remote sensor data (Anderson et al. 1976). The USGS derived land use and cover by conventional interpreta- Methodology tion of high-altitude color-infrared photographs onto The Illinois Geographic Information System (IGIS) base maps at a scale of 1:250,000. was the primary tool used to perform the work de- Data for the study area were extracted from this data Riparian Wildlife Habitat Evaluation 641

Figure 1. Stream names (with Ͼ10 mi2 drainage) for the Salt Fork and Middle Fork branches of the Vermilion River, along with a 300-m buffer, as derived from 1:100,000 DLG data. 642 L. R. Iverson and others

Table 1. Summary of data sets used in this studya

Scale/ Data Sets Date resolution Type Source Stream Network DLG hydrology Varies 1:100,000 Arc coverage USGS 1:24,000 base map series Land cover LUDA 1978, 1881 1:250,000 Arc coverage Aerial photography NHAP 1981–1983 1:24,000 Arc coverage Aerial photography Landsat imagery 1988 30 meters Erdas ﬁle Landsat thematic mapper NWI 1981, 1983 1:24,000 Arc coverage Aerial photography, 58,000-altitude INAI Varies 1:24,000 Arc coverage Illinois Department of Conservation Other Drainage basin boundary Varies 1:24,000 Arc coverage USGS WRD drainage basin area ﬁles Public land survey sections Varies 1:24,000 Arc coverage USGS 7.5- and 15-minute base maps aSee text for data descriptions.

Table 2. Land use and cover categories for LUDA and NHAP dataa

LUDA/NHAP Level I code Corresponding LUDA/NHAP category Agriculture 211 Inactive cropland 213 Active cropland 23 Confined feeding operations 24 Other agricultural land 21 Cropland Grassland 212 Active pastureland 172 Landfill sites (closed sanitary) 62 Grassland wetland Forest 411 Riparian vegetation (trees) 413 Deciduous forest with housing 414 Forested wetland (reclaimed strip) 415 Deciduous forest (reclaimed strip) 416 Riparian forest (reclaimed strip) 61 Forested wetland Mixed successional 22 Orchards, nurseries, and horticulture areas 44 Secondary growth (shrubs, etc.) 412 Riparian vegetation (grassland/shrubs) Urban 11 Residential 12 Commercial and services 13 Industrial 14 Transportation, communication, and utilities 15 Industrial and commercial complexes 16 Mixed urban and built-up areas 17 Other urban and built-up lands aLevel II codes used in LUDA are presented as the first two digits, while three digits represent level III codes used in NHAP.

set. They date from 1978 (Peoria 1:250,000 quadran- the National High Altitude Photography (NHAP) pro- gle) and 1981 (Danville 1:250,000 quadrangle). Two gram. Similar to LUDA, the land use and cover infor- levels of classification were used in the study. Level I mation was determined by interpreting high-altitude consists of nine categories, six of which occur in Illinois: color-infrared photographs, only the classes were urban or built-up, agricultural, forest, water, wetland, mapped at a much finer scale (1:24,000 for NHAP vs and barren. Level II further subdivides each of these 1:250,000 for LUDA). The NHAP photographs categories; these classes are presented as the first two (1:20,000 scale; film exposure 1981–1983) were ob- digits in the LUDA codes in Table 2. tained from the US Department of the Interior, EROS National High Altitude Photography program. Land use Data Center in Sioux Falls, South Dakota. Land use and and cover patterns for the study area were derived from cover patterns were interpreted and digitized as de- Riparian Wildlife Habitat Evaluation 643

scribed in detail in Osborne and Wiley (1988). The Analysis of Land Cover by Data Source land use and cover classification system employed con- The USGS LUDA, the NHAP, and the classified sisted of a modified version of the LUDA system Landsat TM coverages describe land cover throughout (Anderson et al. 1976), which is generally considered to the entire Vermilion basin. For purposes of comparison be resource oriented (see codes and categories, Table and analysis, all of these data were overlain with the 2). 300-m buffer of DLG stream segments. The land use Landsat Thematic Mapper. Landsat Thematic Mapper and cover composition of the area within 300 m of each (TM) satellite data were used to assess land cover for a stream and for the entire study area were determined majority of the watershed study area. The satellite col- for each data set, as defined by that data set. lected the data on 27 June, 1988 and the data have a The percentage of area classified as wetlands, natu- spatial resolution of about 30 m ϫ 30 m per pixel. The ral areas, and preserves within a 300-m buffer (600 m configuration used for this study was ERDAS, Inc. im- total width) was also determined for each of the 30 age processing software, running on a Unix worksta- tributaries contained within the Vermilion River basin tion. and for the entire basin. Geographic reference was established to an accuracy of within half a pixel (approximately 15 m) using 60 Analysis of habitat by tributary ground control points. The study area boundary was used to clip the TM data to create a file containing TM Wildlife habitat. Methods were developed to convert data for the study area only. The TM scene used for this information among the data sets into estimates of value study covered all except about 19,000 ha (8%) of the for wildlife habitat. These methods account for differ- eastern portion of the watershed study area. ences in how land use was coded among the data sets. The TM data were classified into land cover types A principle purpose of this project was to identify using unsupervised classification. The resulting clusters areas of riparian wildlife habitat and the quality of those were assessed for their spatial distribution and, with the habitats. Habitat is considered to be species specific aid of aerial photographs and quadrangle maps, as- with regards to physical surroundings and prevailing signed to land cover types. In most situations, many environmental conditions. It is not realistic to attempt clusters were grouped together to represent one land to identify the habitat of every wildlife species (bird, cover type. mammal, amphibian, insect, etc.) that occurs in Illi- National Wetlands Inventory. Wetland data for the nois; nor should it be necessary to do so from a man- Vermilion River basin were taken from the statewide agement perspective. So, we use the term “habitat” in National Wetlands Inventory (NWI), which resides the generic sense as the habitat of a guild of species. within the IGIS. The NWI, developed by the US Fish The concept of guild was introduced by Root (1967) and Wildlife Service (Cowardin et al. 1979), is based on and refers to a group of organisms of the same taxo- aerial photography from an altitude of 58,000 ft. The cene that utilize a resource in a similar manner. photographs were manually interpreted and tran- Selection of guild categories was limited by the land scribed to the USGS 1:24,000 base map series. The use and cover categories that comprised the original photography for the Vermilion basin was taken in the land use and cover classification systems (e.g., Ander- spring and fall of 1981 and 1983. No differentiation son et al. 1967) used in each of the individual databases among wetland types was made for this study. and by the level of resolution of the different data sets. Illinois natural areas inventory and Illinois nature pre- Therefore, our goal was to encompass the largest num- serves. Natural areas and nature preserves data for the ber of wildlife species within the framework of these Vermilion basin were taken from the statewide Illinois data constraints. We selected three fairly general guilds: Natural Areas Inventory (INAI) (White 1978). Data forest, grassland, and mixed successional. The specific taken on field visits were used to digitize boundaries LUDA land use and cover codes that characterize each drawn on 1:24,000-scale quadrangle maps and low-alti- guild are shown in Table 2. tude photography (8000 ft altitude). The areas identi- A general guide in conservation biology is that a fied are considered to have a unique biological or species can be best protected and managed by manag- cultural value to the state but do not reflect any infor- ing the habitat (e.g., Graber and Graber 1976, Block mation on land ownership. The INAI data quite reliably and Brennan 1993, Block et al. 1994, Lindenmayer et depict the state’s best 0.05% of land still existing in a al. 1994, Rich et al. 1994). Thus, using general habitat natural or near-natural condition. Usually, the nature guilds would seem consistent with contemporary natu- preserves, which are now protected by Illinois statute, ral resource management philosophies. Example spe- are a subset to the INAI. cies from eastern Illinois for our three general guilds 644 L. R. Iverson and others

could be the gray squirrel (Sciurus carolinensis) and Swainson’s warbler (Limnothlypis swainsonii) for forest, the 13-lined ground squirrel (Spermophilus tridecemlinea- tus) and grasshopper sparrow (Ammodramus savanna- rum) for grasslands, and white-tailed deer (Odocoileus virginianus) and red-winged blackbird (Agelaius phoeni- ceus) for mixed successional habitats. Habitat rating by guild for LUDA, NHAP, and TM data. Our goal was to calculate tributary and bank-specific wildlife habitat indices from information in the LUDA, NHAP, and TM data sets. The percentage of each land cover type from the three data sets was calculated for each tributary as a percentage of the total area encom- passed by a buffer extending 300 m either side of the stream. This relatively wide buffer distance was used for several reasons: (1) it matches the buffer for another related Illinois data base, the Illinois Streams Informa- tion System, so that general comparisons can be made; (2) it encompasses most of the forest in the state [79% of all forest lands in another watershed in southeastern Illinois fell within 300 m of streams (Iverson 1994)]; and (3) it ensures that the indices developed consider larger blocks of habitat, not just the habitat immedi- ately adjacent to the streams. Percentage of land cover and the subsequent habitat ratings were developed for grassland, forest, and mixed successional wildlife guilds. Development of such a system proceeded under the following constraints: (1) the rating system had to be Figure 2. Theoretical relationships (i.e., species area curves) applicable to land uses within 300 m of a major stream between the area of habitat and the number of species the reach (0–300 m inclusive); (2) the rating system had to area supports for (A) forest and grassland species, and (B) accommodate forest, grassland, and mixed successional mixed successional species. B: the parabolic curve for forest and/or grassland indicates that these land covers contribute wildlife species guilds; and (3) the system needed to be to mixed successional habitats when their respective cover built with the imposed LUDA classes of land types proportions are Յ20% (e.g., they have significant amount of (Table 2). edge); beyond 20% cover, the forest or grassland cover does Once comprehensive and relatively concordant land not contribute to the mixed successional habitat ranking. use and cover categories were generated, a system for valuation of the proportion of riparian land cover within each category along each stream was developed. proportion of each of the major land use and cover After examination of several potential procedures, we categories (Table 2) within each 300-m buffer of a concluded that the following approach would be most tributary. Subsequently, a score of the relative quality of appropriate for comparison. Evaluation of forest and the riparian habitat (within 300 m of the stream) is grassland habitat was straightforward because of consis- calculated for each guild and for each tributary. The tent classification among data types, whereas the mixed scoring system, which ranges from 0 (no habitat) to 5 successional guild was addressed in a separate, more (optimal habitat) in units of one tenth, is based upon complicated fashion. the general ecological principle of species area curves, Valuation of forest and grassland habitat. The procedure that the number of species inhabiting an area of a given uses interval measurements of land use and codes habitat type is a function of the size of the area or them. It assumes that no land uses other than forest amount of habitat. This relationship is generally as- and grassland, the primary land uses for defining and sumed to be logarithmic in form as depicted in Figure characterizing a species guild of interest, are of value or 2A. The score does not depend on the area of individ- importance as wildlife habitat for the guilds in ques- ual patches within the 600-m strip, only the proportions tion. The initial step involves the determination of the of the land covers within that area for each tributary. Riparian Wildlife Habitat Evaluation 645

Therefore, it assumes that habitat value of several land, and a subdominant parabolic-like relationship to smaller fragments is equal to that of a single patch of exist between mixed successional species and both the same size (as found by McCoy and Mushinsky grassland and forest habitat within the range of 0%– 1999). 20% of forest and/or grassland (Figure 2B). In our Figure 2A was used as a model to generate the approach, we also consider the grassland and forest scoring procedures for grassland and forest species habitats within the range of 0%–20% to be of lesser guilds. This scoring scenario assumed that the species value (i.e., half) than equivalent amounts of mixed richness of forest and grassland species increased as a successional habitats. function of the log of the size of the specific forest or This approach is presented by the following equa- grassland habitats. It also assumed that habitats that tion that calculates a potential mixed successional hab- were composed of a minimum of 85% grassland or itat (PMSH): forest were optimal habitats for grassland and forest % PMSH ϭ ͑% mixed successional͒ guild species (i.e., Ն85% received a score of 5). Therefore equation 1 was adopted to calculate the ϩ ͓% grassland ͑Յ20%͒/2͔ habitat index score for the forest and grassland guilds. ϩ ͓% forest ͑Յ20%͒/2͔ (2) ϩ ͔͒ ͑ ء͒ ϭ ͓ ͑ ϩ ͒ ͑ rnd HI log PCT 1 5/log 101 0.15 (1) The HI value for mixed successional species is then where rnd is a function that rounds the habitat index calculated by substituting the percentage of PMSH into (HI) for the grassland or forest guild to the nearest equation 1 for PCT. All other procedures are as previ- tenth, and PCT is the percentage of forest or grassland ously described for equation 1 above. land use (see Table 2 for pertinent land use codes). If Similar to procedures for determining the HI for the PCT is Ͼ85%, this formula will result in an HI value forest and grassland guilds, the proportions of the total greater than 5. In such instances, the rnd function area within 300 m of the stream that are composed of rounds all values back to a maximum HI value of 5. mixed successional, grassland, and forest land use cat- Equation 1 also dictates that the minimum value of HI egories are calculated. If the grassland and/or forest is 0.15, not 0. categories individually comprise Յ20% of the total area Procedure for mixed successional guild. Evaluation of of land, then half of their proportional area is added to mixed successional species is more complex. Although the total proportion comprised of the mixed succes- we used a mixed successional habitat land-use category sional category (see equation 2) in determining the (Table 2) and have assumed a logarithmic relationship percentage of potential mixed successional habitat. between mixed successional species and the area of Equation 2 incorporates most, but not all, of the mixed successional habitat (Figure 2A), forest and preceding concepts associated with the generalized grassland habitats may also be beneficial to the mixed model depicted in Figure 2B into a single and relatively successional habitat guild. The positive effects of the simplistic scoring system. Although simple, the pro- forest and grassland habitats are related to the benefi- posed model provides some interesting and pertinent cial effects of edges (boundaries between habitat types) hypotheses for future research and testing. and land-cover diversity that create appropriate habitats We reiterate that although calculations of HI utilize for this transitional guild. Too much forest or grassland formal equations, the final values only represent our habitat is likely to detract from the overall quality of the best educated guess of the quality of a land category of habitat for mixed successional species. For instance, we a certain area for a particular guild. The approach hypothesize that a range of 0%–20% of grassland outlined above is still a Delphi procedure and should and/or forest could be suitable for many mixed succes- not be regarded as anything more powerful. Unfortu- sional species. Of course, not all areas of grassland and nately, we were unable to field test these concepts or forest habitats (Table 2) provide appropriate areas for results in this study. mixed successional species; however, the peripheral Table 3 provides representative HI values for mixed margins would be expected to be most appropriate. successional habitats that result from different combi- Therefore, grassland and forest habitats should not be nations of mixed successional, grassland, and forest scaled the same (or worth as much from a quality land cover combinations that were generated using the perspective) as an equivalent amount of mixed succes- preceding method. sional habitats. A generalized model of this concept is Comparisons, mapping and analysis of NHAP, LUDA depicted in Figure 2B. In essence, we would expect a and TM data sets. The methodologies described were dominant logarithmic relationship to exist between applied to each of the 30 stream reaches in the Vermil- habitat quality and percentage of mixed successional ion basin for each guild. For each tributary, this proce- 646 L. R. Iverson and others

Table 3. Range of habitat index for mixed 1.7% grassland (0%–12.6%), 2.3% mixed successional successional guild based on proportions of grassland, (0%–11.6%), 77.7% agriculture (45.6%–100%), 1.9% forest, and mixed successional habitats urban (0%–20.4%), and 0.6% water (0%–7.6%). Land cover type (%) For the entire TM study area (minus the 8% land not covered by the TM data), the land-cover percent- Habitat index Mixed ages were similar to that found in the other data sets score successional Forest Grassland (Figure 4b). Forest within the stream buffers for the 0.9 1 0 0 TM data averaged (range) 9.4% (0%–29.3%), 4.7% 2.2 1 5 5 (0%–19.6%) for grassland, 4.0% (0%–8.9%) for mixed 2.9 5 3 10 3.4 7 1 25 successional, 79.5% (0%–99.9%) for agriculture, 2.1% 3.2 10 5 5 (0%–20.5%) for urban, and 0.3% (0%–1.1%) for water. 3.8 20 1 15 The level II LUDA data did not differentiate any 4.2 35 10 5 grassland for this region. For the entire basin and thus 4.3 40 4 4 excluding the possibility of grassland, the percentage of each land cover ranged from 0.1% for mixed successional and water to 2% for forest to 94% for agriculture dure allowed calculation of average habitat rating (Figure 4c). As with the other data sets, there was a wide scores, which were plotted to visually display habitat for variation among the 30 tributary buffers, with an aver- each guild by data set (NHAP, LUDA, and TM). Fur- age (range) forest of 6.0% (0%–22.5%), 0.2% (0%– ther, correlations among data sources were performed 5.1%) for mixed successional, 90.5% (75.2%–100%) to assess relative performance of our rating schemes for agriculture, 2.5% (0%–19.7%) for urban, and 0.3% and the overall methodology. Spearman rank correla- (0%–3.4%) for water. tions (nonparametric) were run to compare the order Overall, trends for the LUDA, NHAP, and TM data of habitat rankings among the 30 tributaries, and Pear- sets show that the percentage of forest, grassland, and son product moment statistics were generated to assess mixed successional land cover was higher in the buffer similarities in actual habitat ratings. than in the basin as a whole (Figure 4). Not surpris- ingly, the reverse was true for agricultural and urban Results and Discussion areas. The percentage of water was very low but tended to be higher in the buffer as compared to the total Analysis of Land Cover by Data Source basin. The substantially higher proportion of forest, We were interested to learn of the inherent differ- grassland, and mixed successional habitat within 300-m ences and similarities among the data sets, apart from of the streams reﬂects the importance of riparian zones the rankings for habitat value. In this section, we report to wildlife in Illinois and the obvious need to manage on the proportions of each tributary’s 300-m buffer, as these valuable resources. Mounting evidence also shows depicted by various data sets, contained in forest, grass- the value of the riparian vegetation in reducing sedi- land, mixed successional, agricultural, urban, wetlands, mentation and nutrient loading to the streams (e.g., and natural areas. For comparison, we also report these Osborne and Wiley 1988; Welsch 1991; Castelle et al. proportions for the entire Vermillion River basin. 1994; Gilliam 1994; Vought et al. 1994; Lull et al. 1995). Assessment of LUDA, NHAP, and TM data sets. The Average forest cover was highest in NHAP (15.6%), land cover percentages for the 300-m buffer around followed by TM (9.4%) and then LUDA (6.0%) data each stream tributary according to the NHAP data are (Figure 4). This trend can be attributed to the decreas- presented in Figure 3. The percent cover in forest, ing resolution of the data, which will fail to capture grassland, or mixed successional varies widely among smaller and smaller forest patches that are common in the 30 tributaries, from 0 to 53%. The average cover this highly fragmented landscape. The importance of percentages for all 30 tributaries and for all six major such small habitat patches to wildlife is, of course, land covers, within the buffer and over the entire basin, species dependent. The TM data had somewhat higher are given in Figure 4 for each of the three data sets. proportions of grassland and mixed successional land The NHAP data set used the level III classiﬁcation cover than did the NHAP coverages (Figure 4). These shown in Table 2. Over the entire basin, land cover differences can be attributed to the manner in which percentages were dominated by agriculture, which oc- land cover was interpreted and coded as well as differ- cupied a total of 91% of the basin (Figures 3 and 4a). ences in resolution. Among streams in the basin, average land cover per- The Pearson product moment correlation among centages (and ranges) were: 15.6% forest (0%–43.1%), land cover categories for the three data sets revealed Riparian Wildlife Habitat Evaluation 647

Figure 3. Percentage of forest, grassland, and mixed successional land within the 300-m buffers, by tributary according to the NHAP data. The remaining area is agricultural, urban, or water. that the percentage of forested, agricultural, and urban (0.75%). Nature preserves are found along only two land cover types was highly similar for all three data sets streams, the Middle Fork (with 5.06% of the buffer in (r Ͼ 0.6, P Ͻ 0.001, Figure 5). The same was true for nature preserves) and Windfall Creek (25.79%). Based the grassland and mixed successional categories in the on these results, streams rating highest for wetlands NHAP and TM data sets (LUDA had no grassland and and high-quality natural areas include the Middle Fork, only two tributaries with mixed successional land, so no Salt Fork, and Windfall Creek. correlation values were given). The percentages of wa- Analysis of Habitat by Tributary ter were correlated only for LUDA and NHAP (Figure 5). Habitat rating by guild for LUDA, NHAP, and TM data. Assessment of NWI and NAI data sets. A greater per- The habitat ranking scores for LUDA, NHAP, and TM centage of wetlands, natural areas, and nature preserves data are presented in Table 4. There is consistency in are contained within the 300-m buffer of each stream ranking of the streams with high riparian habitat value tributary than are contained in the basin as a whole; across all of the data sets. For instance, the Salt Fork, proportionately four times as many wetlands, three Middle Fork, Glenburn Creek, and Knights Branch all times as much natural area, and two times as much rank high in forest guild habitat value regardless of the nature preserve are found within the buffer than in the land use data set (Table 4). basin. This pattern provides additional evidence that, The LUDA data ranks the Salt Fork the highest for the Vermilion River basin, the most valuable sources (3.57), followed by Collison Branch (3.31) and the for high-quality habitat for wildlife are along stream Middle Fork (3.09). The rest of the streams had ratings riparian zones. Streams that have the largest percent- below 2.8. The grassland guild was not rated due to lack age of wetlands are the Salt Fork (with 14.7% of its of grassland in the LUDA data. Streams with the high- buffer area contained in wetlands), Middle Fork est rating for the mixed successional guild were Colli- (12.4%), and Windfall Creek (11.3%). Streams that son Branch (2.61) and the Middle Fork (2.41). All contain the most natural areas are the Middle Fork other streams had ratings of 2.10 or lower. (1.5% of the buffer area in natural areas), the Saline With NHAP data, the forest guild rates the Salt Fork Branch Drainage Ditch (0.86%), and the Salt Fork at 4.25, Bean Creek at 4.12, and Middle Fork at 4.04 648 L. R. Iverson and others

Figure 4. Average percentage of land cover for stream buffers and the entire Vermilion basin, according to the three data sets. Riparian Wildlife Habitat Evaluation 649

Figure 5. Pearson product moment correlations of percentage of forest, grassland, mixed successional land, agriculture, urban land, and water among three data sources. LUDA data had insufﬁcient grassland or mixed successional data for statistical correlation.

(Table 4). All other streams had ratings below 4. The the higher quality forest habitat was restricted to a few grassland guild listed Collison Branch with the highest streams, usually the larger, high-order streams. There rating (2.98), followed by Feather Creek (2.39) and was generally a good correspondence among data sets Knights Branch (2.36). All other streams had ratings for their ratings. below 2. The mixed successional guild rated Knights Contrary to the habitat ratings for forest species, the Branch at 3.56 and Collison Branch at 3.03. All other grassland habitats tended to be highest on the smaller stream ratings were lower than 3. streams (Table 4). Due to difﬁculty in the ability of the The TM data set listed the top-rated stream for the NHAP data sets to clearly distinguish grassland from forest guild as Collison Branch (3.85), followed by Bean mixed successional habitats, streams received generally Creek (3.82), Salt Fork (3.75), and Middle Fork (3.53). lower habitat ratings for NHAP compared to the TM All other streams were rated 3 or lower (note, however, data sets. The highest NHAP stream rating was 2.98 for that four streams are not represented or only partially Collison Branch. TM rated Collison Branch, Knights represented in the TM data). The top two streams in Branch, and the Salt Fork as the highest with ratings of the grassland guild were Collison Branch (3.43) and 3.43, 3.31, and 2.77, respectively. LUDA had no grass- Knights Branch (3.31). The other streams were lower land identiﬁed for the basin, so habitat scores were not than 3. The mixed successional guild rated Knights generated. Overall, habitat ratings for grassland species Branch the highest (3.55), followed by Collison Branch in the Vermilion River basin were quite low. (3.15), and Stoney Creek (3.10). These top three For the mixed successional guild, ratings were varied streams match those rated highest for the NHAP data over the three data sets, with streams generally being set with respect to habitat for mixed successional species. rated higher according to the TM data set (Table 4). Comparison, mapping, and analysis of LUDA, NHAP, The LUDA data set rated Collison Branch (2.61) and and TM data sets. Visual representations of each data the Middle Fork (2.41) the highest. Many streams, in- set’s habitat ratings by guild for each of 30 tributaries cluding the Salt Fork, had zero ratings from LUDA data were generated from the data in Table 4. An example (which by default equaled 0.15). The NHAP data set using TM data for ranking habitat for the forest species rated Knights Branch (3.56) and Collison Branch is given in Figure 6. For the forest guild, ratings were (3.03) the highest. It also had several streams rated fairly similar among the three data sets, with NHAP between 2.25 and 3; the Middle Fork was rated 2.52 and data rating streams slightly higher (Table 4). The Mid- the Salt Fork 1.39. Other streams rating fairly high were dle Fork, with scores ranging from 3.09 to 4.04, Salt Collison Branch (3.03), Stoney Creek (2.99), and Blue- Fork (3.57–4.25), and Collison Branch (3.31–3.96) grass Creek (3.03). The TM data again rated Knights consistently received the highest values. Many of the Branch (3.55), followed by Collison Branch (3.15), as remaining streams had much lower ratings. In general, the highest rated streams for mixed successional habi- 650 L. R. Iverson and others

Table 4. Habitat rating (0–5 scale) by tributary for LUDA, NHAP, and TM data setsa

Forest Grassland Mixed successional Tributary LUDA NHAP TM NHAP TM LUDA NHAP TM Salt Fork 3.57 4.25 3.75 1.66 2.77 0.15 1.39 3.09 Jordan Creek 0.15 2.93 3.03 0.98 1.75 0.15 2.34 2.82 Stoney Creek 0.99 3.04 2.73 1.64 1.94 0.65 2.99 3.1 Feather Creek 0.15 1.43 0.96 2.39 2.59 0.15 2.07 2.21 Stoney Creek tributary 3 0.15 0.15 0.15 0.15 0.18 0.15 0.15 0.16 Stoney Creek tributary 5 0.15 0.15 0.15 0.15 0.43 0.15 0.15 0.3 Olive Branch 1.21 2.66 2.28 0.86 1.81 0.81 2.1 2.36 Salt Fork Tributary 5 0.15 2.01 0.9 0.15 1.87 0.15 2.01 2.1 Salt Fork Tributary 7 0.15 0.53 0.4 0.15 0.65 2.10 0.36 0.63 Saline Branch drainage ditch 1.71 2.69 2.49 1.75 1.24 1.46 2.47 2.78 Saline Branch tributary 1 0.15 0.15 0.15 0.15 0.16 0.15 0.15 0.16 Saline Branch tributary 3 0.15 0.15 0.15 0.15 0.17 0.15 0.15 0.16 Spoon River 0.15 0.59 0.48 0.43 0.65 0.15 0.85 0.9 Spoon River tributary 0.15 0.48 0.16 0.15 1.81 0.15 1.34 1.6 Upper Salt Fork drainage ditch 0.15 0.60 0.44 1.98 1.69 0.15 1.5 1.31 Union drainage ditch 0.15 0.15 0.17 0.15 0.98 0.15 1.19 1.09 Flatville drainage ditch 0.15 0.15 0.15 0.15 0.78 0.15 0.15 0.51 Upper Salt Fork drainage ditch tributary 0.15 0.15 0.15 0.15 0.49 0.15 0.15 0.33 Middle Fork 3.09 4.04 3.53b 0.72 2.48b 2.41 2.03 2.83b Glenburn Creek 2.28 3.78 2.01b 0.41 2.06b 1.67 2.48 2.67b Windfall Creek 2.74 3.60 —b 1.88 —b 2.08 1.33 —b Collison Branch 3.31 3.96 3.85 2.98 3.43 2.61 3.03 3.15 Knights Branch 1.49 3.12 2.92 2.36 3.31 1.02 3.56 3.55 Bean Creek 2.32 4.12 3.82b 0.15 2.52b 1.70 0.15 2.81b Bluegrass Creek 0.15 3.13 2.57 1.95 2.48 0.15 2.8 3.03 Buck Creek 0.15 2.28 2.17 0.15 2.15 0.15 2.78 2.66 Middle Fork Vermilion River tributary 0.15 1.34 0.67 0.15 2.13 0.15 2.52 2.39 Sugar Creek 0.15 1.72 1.57 0.15 1.18 0.15 2.38 2.37 Praire Creek 0.15 1.76 1.7 0.15 1.3 0.15 2.25 2.21 Wall Town drainage ditch 0.15 1.56 1.53 0.15 1.14 0.15 2.25 2.21 aLUDA did not record any grasslands, so no score is given there. bThese streams have some TM data missing, which accounts for some of the large discrepancies in habitat ranking. tats. The Salt Fork rated much higher than in the ship for the grassland class. All comparisons involving NHAP data set with a rating of 3.09. LUDA were lower, due to lower resolution data and due From analysis of these three data sets, it is clear that to little or no area in the mixed successional or grass- the streams with the best potential overall wildlife hab- land types. The LUDA data are inadequate for this type itat include the Salt Fork, Middle Fork, Collison of analysis, whereas the TM provides a feasible alterna- Branch, Knights Branch, and Bean Creek. It is also tive to the very labor-intensive effort of analyzing high- evident that successional and grassland land cover cat- resolution aerial photographs (as was done with NHAP egories, as compared to forest, produced wildlife habi- data). tat ratings that were more varied and less reliable due to inconsistent classiﬁcation, inadequate spatial resolution (with LUDA), small areas, and small average patch size. Conclusions The habitat rating for the three land cover sources (LUDA, NHAP, and TM) were compared using a Spear- 1. The method presented can successfully produce man rank correlation test (Figure 7a) and a Pearson wildlife habitat rankings by tributary. The three product moment correlation test (Figure 7b). Both data sets (TM, NHAP, LUDA) compared favorably correlation tests revealed similar results. The forest rat- among each other when analyzed at the stream ings among the LUDA, NHAP, and TM data sets were level for forest habitat. NHAP and TM compared all highly correlated. Habitat rankings generated using favorably for grassland and mixed successional hab- NHAP and TM data sets were highly correlated for itats, but inadequate spatial resolution and classiﬁ- mixed successional species as well, with a lower relation- cation inconsistency of the LUDA data was respon- Riparian Wildlife Habitat Evaluation 651

Figure 6. Wildlife habitat ratings mapped for forest species based on TM data.

sible for less than satisfactory results for that data few days. It could also be easily repeated every set. decade or so for temporal comparisons. Hewitt 2. The method can inexpensively and consistently be (1990) also found TM to be useful in inventorying applied to large areas. TM data were shown to riparian forests. NHAP also provided very good provide very good information for this purpose. information, but use of those data are not practical TM, and other readily available satellite data, once on a statewide or larger basis because of the manual they are classiﬁed, can be easily processed for this interpretation that is involved. Possibly, automated ranking scheme. Many states are producing TM- processing on other data, such as digital or- based land cover maps for their gap analysis pro- thophoto quads or other satellite data may be fea- grams (e.g., Scott et al. 1993); these could feed sible and warrants further study. LUDA data were right into this scheme. With suitable GIS hardware acceptable for forest guilds, but the coarser resolu- and software, adequate land cover data, and a tion, lack of grassland and mixed successional hab- stream network database, the entire process could itat classes, and one-time date (late 1970s) limit be performed over an entire state in a matter of a their usefulness in statewide assessments. 652 L. R. Iverson and others

Figure 7. (a) Spearman rank and (b) Pearson product moment correlations among habitat quality ranks for the three databases. No data existed for LUDA grasslands, resulting in 0.0 correlations.

3. Streams within the Vermilion River basin were There is a need for continuing research to imple- rated for habitat value for guilds using forests, ment the work on other basins across the country grasslands, and mixed successional habitat. In gen- and at a greater level of objective precision, while eral, for the forest and mixed successional habitat, minimizing effort. In our results we have tried to the larger the stream and the higher the hierarchi- take into account the many difficulties in rating the cal order, the better quality of habitats for wildlife. value of an area for wildlife habitat and to provide For example, the Salt Fork and Middle Fork rate a useful index for managers and other researchers. higher than the tributaries that flow into them. For Nonetheless, it should be remembered that these grassland habitat, the roles are reversed where the data and the scoring system are generalized to ma- smallest tributaries (low in hierarchical order) have jor wildlife guilds and are based on the best esti- the highest habitat rankings. mates and experiences of the ecologists involved. 4. The results of this project provide a sound founda- Optimally, future work would entail field testing tion for rating riparian habitats across large areas. and perhaps modification of the method as more Riparian Wildlife Habitat Evaluation 653

information is obtained and synthesized. The more Graber, J. W., and R. R. Graber. 1976. Environmental evalua- information available on specific habitat require- tions using birds and their habitats. Biological Notes No. 97, ments of a target species or guilds, the more accu- Illinois Natural History Survey, Urbana, Illinois. rate a relationship can be developed between the Hanson, J. S., G. P. Malanson, and M. P. Armstrong. 1990. Landscape fragmentation and dispersal in a model of ripar- land-cover data and habitat quality. ian forest dynamics. Ecological Modelling 49:277–296. 5. Information presented here from this Illinois study Harris, L. D., and J. Scheck. 1991. From implications to ap- should provide resource managers and policy-mak- plications: the dispersal corridor principle applied to the ers with a mechanism to identify areas what types of conservation of biological diversity. Pages 189–220 in D. A. riparian habitats may yet exist and to make prelim- Saunders and J. Hobbs (eds.), Nature conservation 2: The inary assessments of the quality of the area as wild- role of corridors. Surrey Beatty, Chipping Norton, Australia. life habitat. Such information should be important Havera, S. P., and L. B. Suloway. 1994. Wetlands. Pages 87–151 to the endangered species and heritage programs in Illinois Department of Energy and Natural Resources. throughout the country, as well as scientists and The changing Illinois environment: Critical resources, Vol- ume 3. Illinois Department of Energy and Natural Re- land managers interested in monitoring or improv- sources, Springfield, Illinois. Technical report ILENR/RE- ing wildlife habitat in their locations. EA-94/05. Hewitt, M. A., III. 1990. Synoptic inventory of riparian ecosystems: The utility of Landsat Thematic Mapper data. Forest Acknowledgments Ecology and Management 33/34:605–620. This paper is dedicated to the late Dr. Lewis Os- Illinois Wildlife Habitat Commission. 1985. The crisis of wildlife habitat in Illinois today. Illinois Wildlife Habitat Com- borne, whose life was cut short in the midst of a highly mission Report 1984–1985, Springfield, 26 pp. productive career. This paper would not have been Iverson, L. R. 1988. Land-use changes in Illinois, USA: The possible without his efforts. Thanks to Peter B. Bayley influence of landscape attributes on current and historic for statistical advice and to Doug Johnston, Robert use. Landscape Ecology 2:45–61. Szafoni, Charles Scott, Mary Buchanan, and anony- Iverson, L. R. 1994. Forest resource trends in Illinois. Erigenia mous reviewers for their instructive comments. The 13:4–23. authors are indebted to the Illinois Department of Iverson, L. R., and P. G. Risser. 1987. Analyzing long-term Energy and Natural Resources (now Illinois Depart- changes in vegetation with geographic information system ment of Natural Resources) for their support for this and remotely sensed data. Advances in Space Research 7:183– study. 194. Iverson, L. R., R. L. Oliver, D. P. Tucker, P. G. Risser, C. D. Burnett, and R. G. Rayburn. 1989. The forest resources of Illinois: An atlas and analysis of spatial and temporal trends. Literature Cited Illinois Natural History Survey Special Publication 11, 181 pp. Anderson, J. R., E. E. Hardy, J. T. Roach, and R. E. Witmer. 1976. A land use and land cover classification system for use Klopatek, J. M., R. J. Olson, C. J. Emerson, and J. L. Jones. with remote sensor data. US Geological Survey Professional 1979. Land-use conflicts with natural vegetation in the Paper 964, 28 pp. United States. Environmental Conservation 6:191–200. Block, W. M., and L. A. Brennan. 1993. The habitat concept in Lindenmayer, D. B., R. B. Cunningham, C. F. Donnelly, B. E. ornithology. Pages 35–91 in D. M. Power (ed.), Current Triggs, and M. Belvedere. 1994. Factors influencing the Ornithology. Plenum Press, New York. occurrence of mammals in retained linear strips (wildlife corridors) and contiguous stands of montane ash forest in Block, W. M., M. L. Morrison, J. Verner, and P. N. Manley. the Central Highlands of Victoria, southeastern Australia. 1994. Assessing wildlife-habitat-relationships models: A case Forest Ecology and Management 67:113–133. study with California oak woodlands. Wildlife Society Bulletin 22:549–561. Lull, K. J., J. A. Tindall, and D. F. Potts. 1995. Assessing Castelle, A. J., A. W. Johnson, and C. Conolly. 1994. Wetland nonpoint-source pollution risk. Journal of Forestry 93:35–40. and stream buffer size requirements—a review. Journal of McCoy, E. D. and H. R. Mushinsky. 1999. Habitat fragmenta- Environmental Quality 23:878–882. tion and the abundances of vertebrates in the Florida scrub. Cowardin, L. M., V. Carter, F. Colet, and E. La Roe. 1979. Ecology 80:2526–2538. Classification of wetlands and deepwater habitats of the Naiman, R. J., H. Decamps, and M. Pollock. 1993. The role of United States. Office of Biological Services, US Fish and riparian corridors in maintaining regional biodiversity. Eco- Wildlife Service, Washington, DC, 103 pp. logical Applications 3:209–212. Forman, R. T. T. 1995. Land mosaics. Cambridge University Neely, R. D., and C. G. Heister (compilers). 1987. The natural Press, Cambridge, UK, 632 pp. resources of Illinois: introduction and guide. Illinois Natu- Gilliam, J. W. 1994. Riparian wetlands and water quality. Jour- ral History Survey Special Publication 6, 224 pp. nal of Environmental Quality 23:896–900. Osborne, L. L., and M. J. Wiley. 1988. Empirical relationships 654 L. R. Iverson and others

between land use/cover and stream water quality in a mid- Tazik, P. P., L. L. Osborne, and D. Szafoni. 1991. Nonwoody western agricultural watershed and their application to en- vascular plants in the streams of Champaign County, Illi- vironmental planning. Journal of Environmental Management nois. Transactions of the Illinois State Academy of Science 84:113– 26:9–27. 124. Osborne, L. L., P. Bayley, D. Dowling, R. Larimore, C. Nixon, US Department of Interior and US Geological Survey. 1989. J. Peterson, D. Szafoni, and D. Wood. 1991. The fishes of Digital line graph from 1:100,000 scale maps; data users Champaign County. Final report F-76-R to Illinois Depart- guide 2. Reston, Virginia. ment of Conservation and US Fish and Wildlife Service. Illinois Natural History Survey, Center for Aquatic Ecology Vought, L. B. M., J. Dahl, C. L. Pedersen, and J. O. Lacours- Technical Report 91/5, Champaign, 212 pp. iere. 1994. Nutrient retention in riparian ecotones. Ambio 23:342–348. President’s Council on Environmental Quality. 1978. Environ- mental quality–1978. The 9th annual report of the Council Welsch, D. J. 1991. Riparian forest buffers: function and de- on Environmental Quality. President’s Council on Environ- sign for protection and enhancement of water resources. mental Quality, Washington, DC, 599 pp. Northeastern Area State and Private Forestry, USDA Forest Service, Radnor, Pennsylvania. NA-PR-07-91. Rich, A. C., D. S. Dobkin, and L. J. Niles. 1994. Defining forest fragmentation by corridor width: The influence of narrow White, J. 1978. Illinois natural areas inventory technical forest-dividing corridors on forest-nesting birds in southern report. Volume I, Survey methods and results. Illinois New Jersey. Conservation Biology 8:1109–1121. Department of Conservation, Springfield; University of Root, R. B. 1967. The niche exploitation pattern of the blue- Illinois at Urbana-Champaign, Department of Landscape gray gnatcatcher. Ecological Monographs 37:317–350. Architecture; and Natural Land Institute, Rockford, Illi- nois, 426 pp. Scott, J. M., F. W. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves, H. Anderson, S. Caicco, F. D. D’Erchia, T. C. Ed- Wiley, M. J., L. L. Osborne, and R. W. Larimore. 1990. Lon- wards, J. Ulliman, and R. G. Wright. 1993. Gap analysis: A gitudinal structure of an agricultural prairie river system geographic approach to protection of biological diversity. and its relationship to current stream ecosystem theory. Wildlife Monographs 123:1–41. Canadian Journal of Fisheries and Aquatic Sciences 47:373–384.